MIT study finds humans struggle when partnered with RL agents

The researchers wanted to see if a reinforcement learning program that outperforms humans could become a reliable coworker to humans.